If the digital world were a physical place, it might resemble something like an infinite gallery wall.
We’ve snapped so many shots of ourselves since our phones have come equipped with a camera that the Oxford English Dictionary made “selfie” its word of the year in 2013.
But as our technology has gotten smarter, it’s evolved from the passive pages of our photo albums to the friend actively flipping through those memories with us, recalling the places and names found within.
Such sentience is provided through facial recognition technology—a group of algorithms allowing computers to map out and remember our distinctive facial features with pinpoint aplomb.
The programs have been used for everything from adding fun Snapchat filters to “flipping your gender” on popular Facebook apps.
But it’s not all fun and games.
Officials are praising law enforcement offices for the increased use of facial recognition systems to cut down on identification fraud and help solve a number of other crimes.
The problem, however, is that the algorithms exist virtually free of regulation, and many faces in the crowd—including those who work on developing the programs—are worried what may happen if it’s allowed to grow unchecked.
With its infinite combinations, the human face has long been a source of fascination for everyone from artists to anthropologists, but it’s only recently that we’ve been able to start unlocking the mysteries of its shape.
Artificial intelligence pioneer Woody Bledsoe—with a team of top scientists and modern computers—led the first testing of facial recognition technology as early as 1964.
Aided by a huge book of mug shots, the group was able to extract a list of 20 key measurements needed to map out a face, including the width of one’s mouth, the distance from pupil to pupil, and the exact point of a widow’s peak.
When entered into a computer—and once the pictures were adjusted to account for variation in angles—the calculations were able to identify a subject with startling accuracy. (After the continued success of a later experiment, Bledsoe protégé Peter Hart was reportedly overheard exclaiming, “It really worked!”)
Many of those foundational measurements are still used in facial recognition technology. Today, our powerful machines can take even more dimensions into account, making Snapchat’s funny filters at least as sophisticated as those early experiments.
Still, in order to attach those features to a specific individual, even the smartest computer must first have an idea of whom it’s looking for.
Databases are integral for facial recognition programs to produce a positive ID. And while Bledsoe and company utilized mug shots, many of today’s systems—particularly those used by law enforcement and other government agencies—rely on a different source: the Department of Motor Vehicles.
(Don’t) Put on a Happy Face
Although driver’s licenses may be falling out of favor with today’s teenagers, they’re still incredibly popular. Those who choose not to drive still have a state ID.
As such, the DMV has become a great collector of portraits—and a natural place for facial recognition programs to begin their database scans.
As of 2016, 21 states have allowed law enforcement agencies, from local police to the FBI, access to DMV databases to investigate crimes.
Resembling something of a virtual police lineup, law enforcement officials use the massive backlog of photos to check against the specific biometrics of a particular suspect, bolstered by the technology’s reach and precision.
The technique is currently utilized in at least 41 states to help fetter out driver’s license fraud. And as of 2016, 21 states have allowed law enforcement agencies ranging from local police to the FBI access to the databases to investigate other crimes.
Comparisons will become even easier for computers once every state is in compliance with issuing REAL IDs by 2020. The enhanced identification cards are federally required to be compatible with facial recognition technology—and a big reason many DMVs have issued a “no smile” rule for their headshots.
But even without those considerations in place, it’s already proven an effective enforcement tool. In New York alone, the facial scans were credited with helping police make more than 4,000 arrests between 2010 and 2017, and the state’s DMV issued more than 16,000 administrative actions as a result.
Still, any tool is only as good—or bad—as the person who’s wielding it, and many critics of the technology fear not enough rules are in place to ensure it can’t be exploited.
The widespread use of facial recognition programs has long been the source of heated debate, with some of the technology’s most vocal critics being the very people working on its development.
Last month, Microsoft President Brad Smith added more fuel to that fire, penning a particularly prescient blog post on the subject.
Facial scan systems, he said, could be a potent tool for good, helping reunite families with missing children or stop criminals in their tracks. But the tech could just as easily serve an insidious purpose, spying on the nation and giving everyone from the government to data-hungry corporations insight into exactly where we are, where we’ve been, and what we’re doing. Making matters worse, he said, is that the technology is still far from perfect, which could lead to any number of wrongful accusations or consequences.
To help ensure the programs remain benevolent—at least in their intended purpose—Smith and a growing chorus of others have suggested ceding their designated usage to the federal government. Regulations put in place by a democratically elected body, they argue, are the fairest way to balance the potential for such extremes.
Indeed, Smith and Microsoft have called on Congress to begin examining tough questions on who, exactly, should have final say on such a powerful tool, and just how far that authority should stretch.
But before we begin facing the difficulties on how our visages can be used, we may want to take a long, hard look at ourselves.